• slaughterhouse@lemmy.zip
    link
    fedilink
    arrow-up
    2
    ·
    19 days ago

    LLMs are really good at pattern recognition, making them an excellent tool for scientists and medical researchers.

    • QueenMidna@lemmy.ca
      link
      fedilink
      English
      arrow-up
      13
      ·
      19 days ago

      They can’t infer shit. They don’t recognize numbers as actual numbers, just tokens. They can only repeat what they’ve seen before.

      Combine that with the mathematical certainty of hallucinations and that makes a dangerous combination for scientists and medical researchers.